
As AI models become available to everyone, the advantage for enterprises is shifting to something no model providers can supply: the context that makes AI useful to a specific business. Elastic{ON} Mumbai took on that challenge, bringing together the people building and running AI in India’s enterprises to work out how to make their data count.
The Elastic{ON} Tour 2026/2027 stopped in Mumbai on September 30, bringing developers, SREs, IT and security leaders and practitioners to JW Marriott Mumbai Sahar under the theme ‘Forge the Future’. Across keynotes, live demos, customer and partner sessions, and a hackathon awards ceremony, one idea kept returning: context is what turns a widely available model into a lasting business advantage.
Context is your moat
Ash Kulkarni, CEO of Elastic, who grew up a short distance away in Vile Parle, opened by tying the company’s name to the city. “Mumbai has stretched, it has evolved, it has transformed. It has never broken,” he said.
That adaptability matters, he argued, because AI is moving faster than any earlier technology shift, and the models driving it share a blind spot. “None of these models have any understanding of your business. They don’t understand your data,” Kulkarni said. “The bottleneck is the context.”
Every industry now faces startups trading human capital for token capital. What they cannot copy, Kulkarni said, is a company’s accumulated learnings, experience, and data on how best to serve its customers and partners. “That context is your moat.”
The challenge is that context is scattered. Customer information alone sits across CRM, entitlement, and ticketing systems. Kulkarni’s answer is a bridge that connects it all, securely and with privacy built in, to any model. The result, he said, is “actual intelligence”: the reasoning power of the model combined with the context of the business. Delivering that has meant evolving Elastic’s core engine.
Making context affordable
Ajay Nair, GM, Elasticsearch and Platform, Elastic, picked up the thread. “The model’s only half the equation,” he said. An agent that does not know a customer has raised ten support tickets might still send them a renewal pitch, and that will not go well.
Nair said most teams make their models rebuild context from scratch every time a question is asked, an approach he called “top-down” context that burns tokens on every query. The more efficient pattern is “bottom-up”: deriving relationships, metadata, and risk factors once, keeping them fresh, and handing models a ready-made picture.
That is the idea behind the Elastic Context Engine, which Nair announced on stage and which is now in private preview.
Context at work in operations and security
The same foundation runs through Elastic’s observability and security solutions. Vinay Chandrasekhar, Director of Product Management, Observability, Elastic, demonstrated Elastic nightshift AI SRE, an AI agent that surfaces problems no alert was set up to catch, tests several root-cause hypotheses in parallel, and shows every step of its reasoning. Kulkarni noted that it gets better at investigating “within the context of your environment”, adding: “It’s not generic.” Elastic has since announced nightshift in private preview on Elastic Observability Serverless.
James Spiteri, Senior Director of Product Management, Elastic, argued that “the SIEM has never mattered more”. He previewed what has since launched as AlertZero, the agentic layer built into Elastic Security. Specialised groups of agents, called Watches, triage alerts, hunt for threats, and investigate incidents, learning from how a team already works, while analysts decide which actions need their approval. AlertZero is now available in technical preview across Elastic Cloud, self-managed, and air-gapped environments.
A day built with the community
Beyond the keynotes, customers carried much of the program. On the morning panel, moderated by Ravindra Ramnani, Head of Field Engineering at Elastic India, leaders from Axis Bank, Standard Chartered Bank, and Easebuzz debated whether enterprises can observe, audit, and explain what their AI agents do. Nykaa joined AWS to describe how its SREs now investigate incidents by querying logs conversationally with Elastic AI Assistant. Afternoon breakouts featured Network18 on AI governance, Qualys on closing the SOC loop with agentic AI, and Standard Chartered Bank on standardising observability across 40 countries, alongside community sessions from Iron Mountain, Jolera, and Sprinklr.
Partners showed how they ground AI in enterprise context. Google Cloud explained how its agent tooling and Gemini models work with Elastic retrieval to turn sensitive enterprise data into trusted context without moving it. Microsoft and Elastic walked through OverflowAI, Stack Overflow’s retrieval-first architecture, in which Elasticsearch hybrid retrieval grounds Azure OpenAI in the community’s validated answers.
The community also had its moment on the main stage, as the Forge the Future Hackathon awards recognised the top teams from the Grand Finale held in Bengaluru on September 18 and 19. Frustrated Engineers took first place, followed by Sandboxed in second and Defcon in third. Powered by Elastic and AWS, the hackathon challenged teams to build explainable, production-ready AI agents across search, observability, and security.
Models will change. Context stays.
Kulkarni closed with commitments that put a customer’s context, not any single model, at the centre. Elastic will stay model agnostic, he said, because “there is no winner yet, and there will likely never be a single winner.” It will also run wherever customers do, in the cloud or self-managed.
Models will keep getting better. The context that makes them useful belongs to the business, and Elastic{ON} Mumbai made the case that connecting the two is where the lasting advantage lies.
CTA: Elastic{ON} Mumbai on-demand
https://www.elastic.co/events/elasticon/archive
To know more, visit Make all your data an AI advantage at Elastic{ON}

